home · mobile · calendar · colloquia · 2009-2010 · 

Colloquium - Goldberg

Mining Protein Interaction Networks
Department of Computer Science

In recent years, high-throughput assays for interactions between proteins have provided researchers with abundant graph represented data. Though error-prone, these protein interaction networks (PINs) nevertheless provide deep insights into the function and evolution of proteins.

In this talk I'll describe some of the work we have done to make inferences from PIN data in order to reach a better understanding of protein evolution. First, we found that the prevalence of the de novo acquisition of function by a duplicate gene has been significantly overestimated. Second, we identified putative interaction dynamics (topological trends) for yeast across 300 million years of evolution and mapped the interaction dynamics to biological processes. Third, we identified additional features of evolution absent in current models which contribute to higher clustering, and substantial differences between a model's parameters and the measurements observed as the model runs. This work has led to a new model of protein interaction network evolution which features protein evolution at the domain level. This is joint work with Todd Gibson.

Next, I'll present work that investigates our ability to discover protein complexes from PIN data. We performed a systematic survey of known complexes to better understand which properties are good indicators for protein complexes and what parameter values we can expect to see in complexes within the incomplete data. As part of this survey, we created an algorithm to find the most highly connected subgraph of a graph, a topological property that has not previously been examined within PINs. This is joint work with Suzanne Gallagher.

Finally, I'll briefly describe my efforts to broaden participation in computer science. Together with Dirk Grunwald, Clayton Lewis, Katie Siek, Michael Eisenberg, Alexander Repenning, Roger (Buzz) King, and Shivakant Mishra, we have initiated ECSITE (Engaging Computer Science In Traditional Education), an NSF-funded program that will incorporate computing into existing high school and middle school courses, such as biology, physics, civics, health education, and animation.

Department of Computer Science
University of Colorado Boulder
Boulder, CO 80309-0430 USA
May 5, 2012 (14:13)